Home / Trust
TrustHow we use AI and protect your data
Boards ask two questions before anything else: how do you use AI, and is our data safe? The straight answer, in four parts.
The engine computes the scores, not AI
Fuzzy AHP produces every number, reproducible from the raw answers. A language model drafts the written analysis around figures the engine has already fixed; it never computes, changes or invents a figure, and any sentence whose numbers do not reconcile is rejected.
Your raw data stays with you
With the platform's KPI integration, calculations run inside your own data environment and only the resulting values reach us. No transaction-level records.
Nothing you share trains a model
The hosted model is used for inference only, with the training opt-out set. Strategy text is held only for your engagement, then purged. Where a provider is outside the UK, a UK transfer safeguard applies; the sub-processor register names each one.
Independently tested
Tested by EJN Labs, an independent CREST member firm, in August 2026, with every finding remediated and verified on retest. Planned next: Cyber Essentials Plus, then SOC 2, then ISO 27001; not yet held.
Who sees what
We know who replied
Everything is reported at group level. Each invitation carries a reference unique to the person, so reminders reach the right people; nobody at your company sees an individual's answers, including the CEO.
How many replies the report needs
At least three of your leadership team and five in the layer below. Below either, no report is produced — a line resting on too few responses would not be a fair reading of that group.
What we keep, and for how long
The email addresses you give us are used only to invite and chase, and are deleted with the individual answers sixty days after your readout, or sixty days after the survey closes if no readout has been held within sixty days of the survey closing. Your report's group results, never an individual's answers, are kept for twelve months so a Congruence Map can measure from them, and then deleted. Interview recordings are deleted once the report is written. We also keep your company's anonymised scores, never your company's name, for Congruent's aggregate statistics, unless you opt out.
How a report earns the right to reach your board
Engine figures
Every figure computed by the engine and reproducible from the raw answers.
Narrative checked
Any number in the text without an engine source is flagged, and the report is withheld until it reconciles.
Human sign-off
A named Diagnostic Lead, currently the founder, reviews and signs off every report before it ships.
Auditable record
Each report carries its figure-validation record, so any number can be traced to its source.
For procurement and security teams
The sub-processor register is published in full. The data processing agreement, naming the inference provider, the training opt-out and retention periods, is sent with every order, and on request: Questions: hello@congruentindex.com
Fuzzy AHP
Fuzzy Analytic Hierarchy Process: a peer-reviewed method that turns pairwise judgements into weights. Arithmetic, reproducible, no AI involved.
Inference only
The model generates text from a prompt; it is not given your data to learn from, and the training opt-out is set.
Row-level security
A database control that isolates each client's data so one client can never read another's.
Diagnostic Lead
The role accountable for reviewing and signing off every report; currently the founder.